final

scheduler/scheduler_config.json

@@ -1,10 +1,10 @@
 {
   "_class_name": "VQDiffusionScheduler",
   "_diffusers_version": "0.7.0.dev0",
-  "a_cumulative_end": 9e-06,
-  "a_cumulative_start": 0.99999,
-  "c_cumulative_end": 0.99999,
-  "c_cumulative_start": 9e-06,
-  "num_embed": 4097,
-  "num_train_timesteps": 100
+  "alpha_cum_end": 9e-06,
+  "alpha_cum_start": 0.99999,
+  "gamma_cum_end": 0.99999,
+  "gamma_cum_start": 9e-06,
+  "num_train_timesteps": 100,
+  "num_vec_classes": 4097
 }

transformer/config.json

@@ -1,28 +1,16 @@
 {
   "_class_name": "Transformer2DModel",
   "_diffusers_version": "0.7.0.dev0",
+  "activation_fn": "geglu-approximate",
   "attention_bias": true,
-  "context_dim": 512,
-  "d_head": 88,
-  "depth": 36,
-  "discrete": true,
+  "attention_head_dim": 88,
+  "cross_attention_dim": 512,
   "dropout": 0.0,
-  "ff_layers": [
-    "Linear",
-    "ApproximateGELU",
-    "Linear",
-    "Dropout"
-  ],
-  "height": 32,
   "in_channels": null,
-  "n_heads": 16,
-  "norm_layers": [
-    "AdaLayerNorm",
-    "AdaLayerNorm",
-    "LayerNorm"
-  ],
-  "num_embed": 4097,
+  "norm_num_groups": 32,
+  "num_attention_heads": 16,
   "num_embeds_ada_norm": 100,
-  "num_groups": 32,
-  "width": 32
+  "num_layers": 36,
+  "num_vector_embeds": 4097,
+  "sample_size": 32
 }

transformer/diffusion_pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2b4f3633099b4dcf09ff0187cee3eb6ff3f918fd4f7d1a3dede87ef831b2f277
-size 5437498090
+oid sha256:971c77397048f1473f321bf762c462b181b541b58616721f06d96ec04c00e313
+size 5437498666

vqvae/config.json

@@ -14,7 +14,6 @@
     "DownEncoderBlock2D",
     "AttnDownEncoderBlock2D"
   ],
-  "e_dim": 128,
   "in_channels": 3,
   "latent_channels": 256,
   "layers_per_block": 2,
@@ -27,5 +26,6 @@
     "UpDecoderBlock2D",
     "UpDecoderBlock2D",
     "UpDecoderBlock2D"
-  ]
+  ],
+  "vq_embed_dim": 128
 }